Leading North American Grocer Links Better On-Shelf Availability to $640K in Incremental Sales per Month

• Retailer: Leading North American grocer
• Stores Deployed: Proof of technology phase across eight stores representing two banners
• Focus Areas: On-shelf availability, sales performance, inventory accuracy, store execution
• Key Wins:
  • 6.2% sales uplift, or ~$149K average monthly sales lift per discount store over three months
  • 0.9% sales uplift, or ~$11K average monthly sales lift per supermarket over three months
  • 22% improvement in on-shelf availability (OSA) for the conventional banner, 30% for the discount banner
  • 45% reduction in phantom inventory within 2 months
  • 11% increase in items restocked

A leading North American grocer set out to determine whether continuous visibility into physical store conditions, and more consistent shelf execution based on that intelligence, would improve sales.

The retailer operates multiple grocery formats, including conventional supermarkets and price-sensitive discount stores. The proof of technology phase needed to demonstrate measurable value across both.

On-shelf availability, issue resolution, and revenue impact were the retailer’s primary success measures. Leadership also wanted a credible way to distinguish the effect of better store execution from seasonal patterns and broader market pressure. That was important because comparable stores were experiencing year-over-year sales declines while a distribution disruption impacted fresh product availability.

At the heart of the challenge was a familiar opportunity in retail: creating an accurate, consistent view of what was actually happening inside each store, then turning that ground truth into action.

The retailer’s continuous improvement organization sponsored an eight-store proof of technology phase for Simbe’s Physical AI platform for retail. The platform continuously senses store conditions, transforms observations into SKU-level insights, and helps teams prioritize the actions most likely to improve execution.

Tally, Simbe’s autonomous computer vision robot, served as the primary sensing layer, scanning aisles multiple times per day to create a continuously updated view of availability, pricing and product placement. Four supermarkets and four discount stores participated.

Each location was paired with a comparable control store, and daily point-of-sale and inventory data were used for the evaluation. Year-over-year revenue performance was compared in participating stores against matched controls. This helped account for seasonality, market conditions, and other external factors, providing a more reliable view of the relationship between shelf execution and sales.

Rather than replacing existing store workflows, the retailer introduced physical AI-powered intelligence through processes already familiar to store teams. Grocery managers reviewed prioritized reports, printed task lists, and assigned replenishment work to store associates. Mobile workflows were reserved for a later stage. This adoption model helped teams act quickly and allowed store leaders to see the value firsthand.

Importantly, the platform did more than identify problems. It helped different roles across the organization understand what required attention and then take the appropriate next action. Daily reports helped store teams prioritize out-of-stocks, while district managers and specialists received readiness views for high-sales fresh categories such as bakery, prepared meats, and rotisserie chicken. Teams could quickly see whether they needed to restock an item, correct an inventory balance, review department readiness, or escalate a recurring issue.

For a retailer managing high employee turnover, that consistency was especially valuable: physical AI gave experienced and newer team members a shared, objective view of store conditions and where to focus their time.

“We have a 70% store employee turnover. Tally makes consistent operations easier.”

During the proof of technology phase, participating stores improved shelf conditions, inventory accuracy, and sales performance demonstrating how better in store intelligence translates into measurable business outcomes.

On-shelf availability increased by 22% across the conventional banner and 30% across the discount banner. Stores also restocked 11% more items, while the share of issues resolved within 48 hours rose from 60% to 67%.

The strongest financial result came from the discount banner, where participating stores outperformed matched controls by over 6%. Across all eight locations, the retailer generated approximately $640K in incremental sales per month during the three-month measurement period. Discount stores averaged $149K per store per month, while conventional stores averaged $11K per store per month.

The platform also helped close the gap between what inventory systems indicated and what was actually happening on the shelf. Phantom inventory declined by 45% within two months, helping teams focus replenishment efforts on the issues most likely to affect sales and creating a more accurate foundation for downstream inventory decisions.

Although pricing was a secondary use case, the two paper-tag stores improved accuracy by 69%, further demonstrating additional operational value.

“Pricing reporting is great, it’s helping our team execute better.”

This retailer approached shelf intelligence as a measurable operating program, not a technology experiment. By pairing participating stores with matched controls, leadership could connect more accurate visibility into physical store conditions and stronger shelf execution to improved sales performance.

The program also showed that adoption can begin with familiar workflows, turning continuous store data into focused priorities and actions. Clear reporting gave experienced store leaders and newer team members a more consistent view of what needed attention.

The proof of technology also demonstrated the platform’s compounding value. The same shelf intelligence foundation supported on-shelf availability, replenishment, inventory accuracy, pricing, fresh readiness, and sales performance, creating a foundation for additional use cases over time.

Most importantly, the retailer gained a new way to manage store execution. Instead of periodically auditing stores and reacting to problems after they surfaced, teams could continuously understand physical store conditions, identify where sales were at risk, determine why issues were occurring, and focus action where it would have the greatest impact.

That is the shift Physical AI makes possible: moving from periodic observation to continuously sensing, understanding, and acting on what is happening in the physical store.